Random drift error of laser gyro is one of the main sources of error.Kalman filter algorithm can effectively restrain the random error of laser gyro.But when the model is not accurate,the conventional Kalman filter may easily result in divergent result.Fading Kalman filter can reduce the power of historical information and increase the power of the current measurement information,which may obtain continuous smooth filtering result.The fading Kalman filter was used to estimate the optimal laser gyro drift and a good result was obtained.The result showed that the fading Kalman filter has a smaller residual standard deviation,higher filtering accuracy and stability,and the output sequence is more stable.